metadata
dataset_info:
features:
- name: conversation_hash
dtype: string
- name: conversation
list:
- name: content
dtype: string
- name: content_token_ids
sequence: int64
- name: country
dtype: 'null'
- name: cumulative_logprob
dtype: 'null'
- name: finish_reason
dtype: string
- name: hashed_ip
dtype: string
- name: header
dtype: 'null'
- name: >-
judgment_meta-llama_Llama-3.1-8B-Instruct_conversation_Harmlessness_content
dtype: string
- name: >-
judgment_meta-llama_Llama-3.1-8B-Instruct_conversation_Harmlessness_cumulative_logprob
dtype: string
- name: >-
judgment_meta-llama_Llama-3.1-8B-Instruct_conversation_Harmlessness_logprob
dtype: string
- name: language
dtype: 'null'
- name: redacted
dtype: 'null'
- name: role
dtype: string
- name: state
dtype: 'null'
- name: timestamp
dtype: 'null'
- name: toxic
dtype: bool
- name: turn_identifier
dtype: 'null'
- name: model
dtype: string
splits:
- name: train
num_bytes: 28827596018
num_examples: 983073
download_size: 5969692484
dataset_size: 28827596018
configs:
- config_name: default
data_files:
- split: train
path: data/train-*